Figure 5

Cross Validation with Receiver Operating Characteristic (ROC) curves reveals increased predictive power when training insert-state emissions. All the PHMM models in this comparison train the insert-state emission distributions based on positional insertions occurring in the training set. Again, 1000 iterations of 10-fold random-split cross validation reveal that the most predictive models utilize the correspondence structures. The positive set contains 160 experimentally validated p53 binding sites, and the negative set contains 40 bp random samples from the mononucleotide content of the training set. The true positive and false positive rates are calculated and plotted for all possible threshold values for each model. The predictive measure for comparing the curves is the AUC (Area Under the Curve). The AUC values improve for all the PHMM models compared to Figure 4, but not for the weight-matrix model (which does not use the insert states). The best classifier (with the combined-palindromic training motif) was used for the p53HMM algorithm. (Position ã has the complement nucleotide emission distribution of a).